Comparison · same format, hosted in Europe
Jev made typed decisions a category of its own: ask a question about your data, get a value and a probability back. Xev brings the same System One format to Europe, served from Sweden on hardware AxForge owns, priced in euros, and part of the same account as your chat models, embeddings and GPUs.
At a glance
| Dimension | Xev (AxForge) | Jev (TypeSafe) |
|---|---|---|
| What it is | A decision model: typed answers with calibrated probabilities, never free text. | A decision model: typed answers with calibrated probabilities, never free text. |
| API format | System One: the same request and answer shape; jev-latest accepted as the model name. | System One. |
| Question types | Yes/no · choice · score. | Yes/no · choice · score. |
| Price | €0.022 per 1M input tokens, output free. Launch pricing. | $0.042 per 1M input tokens, output free. |
| Where it runs | Stockholm, Sweden (eu-se-1) on NVIDIA DGX Spark systems AxForge owns and operates. | TypeSafe's cloud. |
| Your data | Zero prompt retention: processed in memory, not written to disk, not logged, never used to train. | Set by TypeSafe's own policies. |
| Confidence threshold | Built in: set min_confidence and unsure answers come back as "decided": false. | Handled in your own code from the returned confidence. |
| Same account also runs | Qwen3.8 27B chat, embeddings, image generation and editing, speech, music, and dedicated GPUs. | TypeSafe's System One models. |
Switching
Your questions, criteria, thresholds and parsing code stay exactly as they are. Only the address and the key move.
# the endpoint
- https://…/systemone # your current System One endpoint
+ https://api.axforge.ai/v1/systemone
# the key
- Authorization: Bearer $TYPESAFE_KEY
+ Authorization: Bearer $AXFORGE_API_KEYThe request body (state, questions,
type, instructions, criteria) is the same. The
Xev docs list every field.
Honest routing
| You're all-in on TypeSafe | You already use TypeSafe's other models and tooling, and one vendor for all of it matters more than where it runs. |
|---|---|
| You bill in dollars | Your spend, budgets and invoices all live in USD. |
Honest routing
| EU data | Your customers' messages, tickets or records should be processed in the EU, and you want to say exactly where: Stockholm, on named machines. |
|---|---|
| Zero retention | Nothing you send may be stored or trained on. Xev processes in memory and keeps only billing metadata. |
| One account for the whole stack | Decisions with Xev, answers with Qwen3.8, search with embeddings, images and speech, and a dedicated DGX Spark when you outgrow serverless. |
| Euro pricing | €0.022 per million input tokens, output free, invoiced in euros, and 3M free tokens every 30 days to start. |
FAQ
Yes. Xev speaks the System One format, so your requests and your answer parsing stay
as they are. Change the endpoint to https://api.axforge.ai/v1/systemone and
the key to your AxForge key.
Xev is €0.022 per million input tokens; Jev is $0.042. Both leave output free. Xev is billed in euros, with no subscription and no minimum.
No, Xev is AxForge's own decision model, built on Qwen3.8 27B and served on our own hardware in Sweden. What it shares with Jev is the format: the same questions in, the same kind of typed, probability-backed answer out.
Yes, and it's an easy way to compare: send the same request to both endpoints and look at the answers. Most teams keep the one that fits their data rules and budget.